Article: Building a Secure MCP Server on AWS for a Million-Company B2B Platform

2026-05-18 · Source: InfoQ · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Cybersecurity & Data Privacy · Depth: Intermediate, extended

Summary

An editorial analyst details the process of building a secure Model Context Protocol (MCP) server on AWS for a B2B intelligence platform managing over one million company profiles. The project aimed to enable LLM clients to interact with the platform via structured requests, such as "find SaaS companies in Germany with 50-200 employees," instead of manual portal queries. The MCP server was treated as a first-class interface, not a mere wrapper, emphasizing its own contracts, security, testing, and operational controls. The architecture leveraged GraphQL on AWS AppSync for data exposure and a Go-based MCP server to translate user requests into narrowly scoped tools. Key design decisions included separating read and write operations, implementing a default-deny approach for mutations, and using layered testing with mocked GraphQL clients and real-system validation via MCP Inspector.

Key takeaway

For AI Engineers integrating LLMs with production systems, prioritize designing MCP servers as first-class interfaces with explicit security controls. You should separate read and write operations at the tool level and enforce a default-deny policy for mutations. Implement robust, layered testing, including mocked clients for unit tests and real-system validation with tools like MCP Inspector, to catch backend-specific failures early and ensure predictable behavior.

Key insights

Treating MCP servers as first-class interfaces with narrow tool contracts enhances security and maintainability for LLM integrations.

Principles

• Separate read and write operations in tool design.
• Adopt a default-deny approach for mutations.
• Narrow tool contracts improve safety and maintainability.

Method

Implement an MCP server with a Go-based tool layer, GraphQL on AWS AppSync, OIDC authentication, and layered testing using mocked clients and MCP Inspector for real-system validation.

In practice

• Use `mcp-go` for GraphQL client integration.
• Implement `allowMutations` flag for mutation tools.
• Capture GraphQL variables in mocked tests.

Topics

• Model Context Protocol
• AWS AppSync
• LLM Integration Security
• Read/Write Separation
• GraphQL

Best for: AI Engineer, MLOps Engineer, Software Engineer

Related on AIssential

• Orchestrating the Platform: A Deep Dive into Model Context Protocol Servers for DevOps and Platform… — Model Context Protocol servers enable AI agents to autonomously manage complex DevOps workflows by standardizing tool interactions.
• How I Connected Claude MCP to an Enterprise Identity Database for Natural Language Querying &… — Combining MCP with dynamic schema discovery and a SQL safety layer enables secure natural language querying of enterprise databases.
• MCP Dev Summit [Day 2] ft AWS, Docker, & Datadog — MCP is evolving to support real-time, interactive, and secure agent experiences, moving beyond simple request-response models.
• MCP as the API for AI‑Native Systems: Security, Orchestration, and Scale — MCP is becoming the API for AI-native applications, requiring secure, orchestrated, and scalable deployments.
• A Taxonomy of Runtime Faults in Model Context Protocol Servers — A new taxonomy empirically categorizes 73 runtime fault types in Model Context Protocol (MCP) servers, enhancing reliability understanding for LLM tool integration.
• Building agents that reach production systems with MCP — MCP standardizes agent-to-system integration, enabling scalable, portable, and context-efficient connections for cloud-based AI agents.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by InfoQ.